In this study, we did not identify strong associations between mid-pregnancy maternal awakening salivary cortisol levels and the occurrence of preterm birth and low birth weight. However, a possible association was observed between high cortisol levels in mid-pregnancy (i.e. ≥90th percentile) and SGA. In the exploratory analyses, this risk of SGA was only increased among female infants, just as the risk of preterm birth, but these analyses relied on small sample sizes, resulting in unstable effect estimates.
Although cortisol has repeatedly been hypothesized as one of the biological intermediates linking prenatal psychosocial stress to adverse birth outcomes [6-8], studies examining the associations between maternal cortisol levels during pregnancy and adverse birth outcomes showed inconsistent results. Reasons for these inconsistencies include differences in biomarkers for cortisol, the timing and extent of biomarker collection throughout the day, and the variability in the timing of assessment during gestation, measuring in early, mid or late pregnancy [39, 40]. Some, mostly small, studies found associations between elevated maternal cortisol levels and the risk of preterm birth in the entire population or in subgroups [19, 41, 42], although publication bias may be an issue here. Other published studies did not observe associations between maternal cortisol levels during pregnancy and preterm birth [20, 43, 44], in line with our results. In a recent meta-analysis among 1606 maternal-fetal dyads , a negative association was observed between maternal salivary cortisol and infant birth weight. However, the risk of low birth weight, a clinically relevant outcome measure, was not assessed. In a large prospective cohort study of 2810 women not included in this meta-analysis, no associations were observed between elevated serum cortisol levels measured in early pregnancy and offspring birth weight . Furthermore, cortisol was not associated with self-reported measures of psychological functioning among pregnant women in several studies [11-13, 23], making it unlikely that cortisol can be classified as biomarker of the mechanistic pathway linking maternal psychosocial problems to preterm birth and birth weight according to the ‘Meet in the Middle’ approach. Alternative markers of stress related to the autonomous nervous system and inflammatory response system may also play a more prominent role in the association between maternal prenatal stress and these birth outcomes [28, 45], and should be taken into account in future studies.
In the same prospective cohort study of 2819 women, an association was observed between high morning cortisol levels (≥90th percentile) and an increased risk of SGA, concordant with our findings. Several potential pathways underlying the associations between elevated maternal cortisol levels and adverse birth outcomes have been suggested, including dysregulation of the fetal autonomic nervous system , vasoconstriction of the uterine artery resulting in reduced uterine blood flow , and stimulation of the production and release of placental CRH . Specifically for SGA, an association with disturbed expression and/or activity of placental 11β-hydroxysteroid dehydrogenase type 2 (11β-HSD2), an enzyme that protects the fetus from high levels of maternal cortisol, has been observed [46-50]. Interestingly, decreased functioning of 11β-HSD2 in SGA infants has only been observed in pregnancies with female fetuses . Recent findings on sex-specific differences in systemic glucocorticoid imbalance  also strengthen the plausibility of our finding that elevated maternal salivary cortisol increases the risk of SGA among female fetuses only. Likewise, previous research indicated that the effects of depression during pregnancy could have a different impact on the two sexes .
Strengths of this study include the nested case-control design with embedding in a large prospective cohort study. Due to this design, we were able to include a relatively large number of cases compared to previous smaller studies [6, 23, 41]. This made it possible to perform stratified analyses for infant sex, which could modify associations between increased cortisol levels and adverse birth outcomes. The design also led to the availability of prospectively collected information on many maternal characteristics, which were used to adjust for potential confounding effects on the associations of interest. Previous studies showed that maternal characteristics, such as parity, smoking behavior, and BMI, could have an influence on maternal cortisol levels [7, 54, 55]. Adjustment for these factors was often not completely or not at all possible in previous studies, resulting in potentially biased effect estimates due to residual confounding.
A limitation of the current study is the collection of only a single awakening saliva sample to determine the cortisol level, whereas multiple measurements during the day or on consecutive days might be preferred. In a previous study validating this approach, however, we showed that a single awakening salivary sample could reliably distinguish between women having normal and elevated cortisol levels . Nevertheless, we were unable to assess the cortisol awakening response and exposure patterns or trajectories throughout pregnancy with a single sample. As the saliva sample was collected in mid-pregnancy, we could not examine the associations between adverse birth outcomes and fetal exposure to elevated cortisol levels in early or late pregnancy. The critical exposure windows for the risks of preterm birth, low birth weight, and SGA are currently unknown [7, 56], but most previous studies collected the biomarkers for cortisol around gestational week 20, comparable to our study. By performing stratified exploratory analyses, we obtained more insight into the potential sex-dependent effects of elevated maternal cortisol levels. Due to small numbers of exposed cases in some of these exploratory analyses, however, we cannot draw firm conclusions on the sex-specific associations. Therefore, these analyses should be repeated in larger studies.